A Simulation Study on the Transient Behavior of a Gasoline Direct Injection Engine under Cold Start Conditions

The cold start process is critical to control the emissions in a gasoline direct injection (GDI) engine. However, the optimization is very challenging due to the transient behavior of the engine cold start. A series of engine simulations using CONVERGE CFD™ were carried out to show the detailed process in the very first firing event of a cold start. The engine operating parameters used in the simulations, such as the transient engine speed and the fuel rail pressure (FRP), came from companion experiments. The cylinder pressure traces from the simulations were compared with experiments to help validate the simulation model. The effects of variation of the transient parameters on in-cylinder mixture distribution and combustion are presented, including the effects of the rapidly changing engine speed, the slowly vaporized fuel due to the cold walls, and the low FRP during the first firing cycle of a 4-cylinder engine. Comparison was also made with non-transient steady state operation. It was shown that the injection-induced tumble ratio in the cylinder varied for different engine speed cases, resulting in a better fuel distribution in the low engine speed case. A relatively high turbulence level during the combustion process was seen in the transient engine speed case that led to strengthened combustion. The fuel tracking from the simulation indicated that about 30% of the fuel remained unreacted at a very late crank angle, in which 8% was in the gas phase and 22% in the liquid phase as wall films. As the FRP increased, the fuel droplets became smaller, and more fuel vaporized before hitting the piston. But the splash and rebound fuel fraction off the piston bowl was even less, resulting in a lower overall gas-phase equivalence ratio in the first firing event for the high FRP case. On the other hand, the fuel distribution was more homogeneous under the high FRP condition due to the high injection velocities. There should be an optimal FRP value, but it was difficult to discern since the peak pressures in different FRP cases were similar and turbulence levels varied in a non-monotonic way, as well, which would lead to cycle-to-cycle variations in the real engine.